This is a Preprint and has not been peer reviewed. The published version of this Preprint is available: https://doi.org/10.1016/j.ecolmodel.2020.109257. This is version 1 of this Preprint.
Downloads
Authors
Abstract
Advances in sensing and computation have accelerated at unprecedented rates and scales, in turn creating new opportunities for natural resources managers to improve adaptive and predictive management practices by coupling large environmental datasets with machine learning (ML). Yet, to date, ML models often remain inaccessible to managers working outside of academic research. To identify challenges preventing natural resources managers from putting ML into practice more broadly, we convened a group of 23 stakeholders (i.e., applied researchers and practitioners) who model and analyze data collected from environmental and agricultural systems. Workshop participants shared many barriers regarding their perceptions of, and experiences with, ML modeling. These barriers emphasized three main areas of concern: ML model transparency, availability of educational resources, and the role of process-based understanding in ML model development. Informed by workshop participant input, we offer recommendations on how the ecological modelling community can overcome key barriers preventing ML model use in natural resources management and advance the profession towards data-driven decision-making.
DOI
https://doi.org/10.31223/X5D01H
Subjects
Agriculture, Computer Sciences, Environmental Education, Environmental Sciences, Environmental Studies, Forest Management, Forest Sciences, Natural Resources and Conservation, Natural Resources Management and Policy, Water Resource Management
Keywords
machine learning, decision-making, natural resources management, stakeholders, decision-support tools, process-based modeling
Dates
Published: 2020-10-22 16:51
Last Updated: 2020-10-22 23:51
License
CC BY Attribution 4.0 International
Additional Metadata
Conflict of interest statement:
None
Data Availability (Reason not available):
There is no data or code associated witht this paper.
There are no comments or no comments have been made public for this article.